[tor-scaling] Performance Experiments + R&D Pipeline Kanban Documents

[Mike Perry]

For our meeting tomorrow, I've worked on a wiki page to sketch out our
first set of live network A/B performance experiments, along with what
metrics they need:
https://trac.torproject.org/projects/tor/wiki/org/roadmaps/CoreTor/PerformanceExperiments

I've done only the four experiments I'm most familiar with in full
detail. I tried to favor metrics from Karsten's first report, where
possible (again: CDFs are preferable to our current website metrics,
because we want to see the full range of performance variance). I also
tried to explain what good results look like and what bad results look
like for each metric.

We still need to fill that page in with in how to tune EWMA, how to test
preemptive circuit building, and any other experiments that can be done
using just torrc and/or consensus parameter tuning on our live network.
If you have ideas in this category, feel free to add them. However,
anything that that requires Tor patches that must be deployed at relays
is out, for this stage.

For items in our long-term plan, I've started adding cards to the R&D
pipeline kanban here (readonly link, since this list is public):
https://storm.torproject.org/shared/k0SG5SfFTGGCsR0poyekmpSb-hLfLynJAUERHyMbGc4

If the phrase "R&D pipeline kanban" makes no sense to you, please read
my proposal to this list that explains it:
https://lists.torproject.org/pipermail/tor-scaling/2019-April/000011.html

The gist of it is that cards on that kanban represent topics in various
stages of development. Cards move from left to right on that kanban as
they move through various stages of completion. Thus, this pipeline
kanban document provides a holistic, high level overview of all of our
performance and scalability projects in each stage of completion, so we
can make priority decisions, keep an eye on capacity, and watch for
bottlenecks and stalls at particular stages of completion.

I'm looking forward to the call tomorrow to hear feedback about all of
this. Remember: We're not yet committing anyone to work on any of this
stuff immediately; we're just trying to make a good game plan to hit the
ground running once we get funding for this work.


-- 
Mike Perry
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: Digital signature
URL: <http://lists.torproject.org/pipermail/tor-scaling/attachments/20190425/a79ad563/attachment.sig>